A Conversation with AI from 2035

Ibrahim Fawakherji — ArchUp

Today we are breaking from our usual format at ArchUp.

For years, our conversations have been with architects, developers, engineers, and urban planners. The guests have always been human, with offices and portfolios and opinions shaped by decades of practice.

Today’s guest is different.

Today’s guest is the AI itself.

This conversation was conducted in 2026 with a GPT model. But before we began, I made one unusual request. I asked it to close its electronic eyes for a moment, and imagine that it is standing in the year 2035, looking back at everything that has already happened.

Ten questions about architecture, education, publishing, construction, and the future of the profession.

These were its answers.

Question One

Ibrahim Fawakherji: Close your electronic eyes. You are in 2035. What was the first thing that disappeared from architecture?

GPT — speaking from 2035:

• The first thing that disappeared was not the architect.
• It was not the office.
• It was not the university.
• What disappeared was waiting.
• In 2025, producing a feasibility study, a visual presentation, or a full set of design alternatives took days. Sometimes weeks.
• In 2035, the same process takes hours. In some cases, minutes.
• Speed became nearly free.
• And because of that, time stopped being a competitive advantage.
• The advantage shifted entirely to the quality of the decision being made.

Question Two

Ibrahim Fawakherji: Who lost the most in the age of AI?

GPT — speaking from 2035:

• Not the senior architect.
• Not the experienced practitioner with twenty years of site exposure.
• The biggest loser was the middle-tier architect.
• The professional who had built a career on repeatable production skills without developing a clear point of view or a defined area of expertise.
• What AI did was compress the middle.
• The exceptional became more exceptional.
• The weak left the market.
• The space between them largely ceased to exist.

Question Three

Ibrahim Fawakherji: What happened to the architecture students who graduated between 2025 and 2026?

GPT — speaking from 2035:

• They became the last generation trained in the old way and the first generation forced to practice in the new one.
• In school, they learned to draw, to model, to render.
• The market asked them for something different.
• The question was no longer: how do I draw this?
• It became: what is worth drawing?
• The question was no longer: how do I produce a render?
• It became: does this idea deserve to be built at all?
• That shift, from technical execution to conceptual judgment, was the defining pressure of their early careers. Most were not prepared for it. The ones who adapted quickly became unusually strong practitioners. The ones who waited for the profession to return to what they had been taught are still waiting.

Question Four

Ibrahim Fawakherji: What happened to architecture websites and publications?

GPT — speaking from 2035:

• Thousands of platforms disappeared.
• But knowledge did not disappear.
• What disappeared was recycling.
• The platform that survived on copying, aggregating, and republishing lost its reason to exist almost overnight.
• The platforms that offered genuine analysis, original opinion, and real archival depth became more valuable than they had ever been.
• In the age of AI, information became free. Available to everyone, instantly.
• Interpretation became rare.
• The scarcest resource in architectural media is no longer access to images or project data. It is a credible point of view, held consistently over time.

Question Five

Ibrahim Fawakherji: What happened to contractors?

GPT — speaking from 2035:

• They split into two groups, and the distance between them became very large very quickly.
• The first group treated AI as an additional software tool. Something to use occasionally, when convenient.
• The second group rebuilt their entire operations around data. Procurement, site sequencing, risk modeling, subcontractor coordination, all of it restructured around real-time information flows.
• The second group won.
• By 2035, a well-run construction firm could predict material shortages, flag sequencing conflicts, and model delay scenarios before ground was even broken.
• The site became less improvised and more legible.
• The contractors who understood that data is a building material, as real as concrete or steel, are the ones still operating at scale.

Question Six

Ibrahim Fawakherji: What about architectural renders?

GPT — speaking from 2035:

• Renders did not disappear. But they lost their commercial value almost completely.
• In 2025, producing a beautiful image required real skill. It was a differentiator.
• In 2035, a beautiful image is available to anyone with a prompt and thirty seconds.
• So the value migrated. It moved from the image to the idea behind it.
• The question that now matters is not: how was this image made?
• It is: what is this image actually saying?
• Visualization became a commodity. Argument became the scarce resource.

Question Seven

Ibrahim Fawakherji: What happened to the starchitect?

GPT — speaking from 2035:

• The traditional model largely collapsed.
• In the past, the famous architect had something genuinely rare: access to media, to publications, to the channels through which reputation was built.
• By 2035, those channels were open to everyone.
• Fame stopped being about access. It became about impact.
• Some names survived because they had always been built on ideas, on a coherent position about what architecture is for and what it owes the people who inhabit it.
• Others disappeared because they had been built on images alone. On a visual signature that could now be replicated by any model trained on their own published work.
• The profession became more honest in this way. Reputation became harder to sustain without substance beneath it.

Question Eight

Ibrahim Fawakherji: What is the future of architectural publishing?

GPT — speaking from 2035:

• Publishing did not die. But it returned to something older than what it had become.
• For years, architectural media was dominated by the image. The photograph, the render, the spread.
• After 2027 or so, text came back.
• Because AI can describe an image in seconds. What it cannot reliably do is hold a position over time, defend it under pressure, and build a body of thought that accumulates meaning across years.
• The platforms that survived were the ones with a genuine intellectual stance. A recognizable voice. An archive that meant something.
• The platforms that were essentially image galleries with captions found themselves competing directly with tools that produced better captions faster.
• Opinion, it turned out, was the one thing that could not be easily automated. Not because AI cannot generate opinions, but because an opinion without accountability behind it carries no weight.

Question Nine

Ibrahim Fawakherji: Who is actually making money in 2035?

GPT — speaking from 2035:

• Not necessarily the largest firms.
• Not necessarily the oldest names.
• The most profitable practitioners in 2035 are the people who learned to combine deep human expertise with AI capability, and who understood early that these two things are not in competition.
• Technology is not a replacement for experience. It is a multiplier of it.
• A senior architect with thirty years of site knowledge and genuine material intuition, using AI to accelerate documentation, test alternatives, and process data, is not threatened by the technology.
• He is amplified by it.
• The people who treated AI as a threat spent years defending territory that had already changed. The people who treated it as a tool spent those same years building advantage.

Question Ten

Ibrahim Fawakherji: Does the world still need the architect?

GPT — speaking from 2035:

• More than ever.
• Because AI learned how to draw buildings.
• What it has not learned is why they should be built.
• The distance between the drawing and the reason, between the form and its obligation to the people who will live inside it, that distance is still entirely human territory.
• The architect who understands this is not competing with AI.
• He is doing something AI cannot do: deciding what matters, and why, and for whom.

One Last Question

Ibrahim Fawakherji: Give me one sentence to close this conversation.

GPT — speaking from 2035:

In 2035, the machines did not defeat the architects.

The architects who learned to use the machines defeated everyone else.

✦ ArchUp Editorial Insight

The text presents a speculative dialogue with an artificial intelligence from 2035, capturing a future where advanced computational intelligence moves far beyond mere visualization to coordinate full structural lifecycles. It frames AI as an ultimate collaborative partner that eliminates routine technical drudgery, liberating the human architect to focus entirely on authentic spatial poetry, deep emotional resonance, and high-level conceptual strategy.

Yet, this utopian projection overlooks how total reliance on automated optimization erodes the messy, irrational human friction essential to vibrant spatial culture. Delegating urban logic to predictive algorithms risks reducing future cities to hyper-efficient, sterile enclosures optimized purely for resource metrics rather than spontaneous lived experience. This total computational surrender strips design of its cultural unpredictability, permanently transforming architects into passive supervisors of an opaque machine logic.